Two-Dimensional Nitrides as Highly Efficient Potential Candidates for CO 2 Capture and Activation†

The performance of novel two-dimensional nitrides in carbon capture and storage (CCS) is analyzed for a broad range of pressure and temperature conditions. Employing an integrated theoretical framework where CO 2 adsorption/desorption rates on the M 2 N (M= Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, W) surface are derived from transition state theory and density functional theory based calculations, the present theoretical simulations consistently predict that, depending on the particular composition, CO 2 can be strongly adsorbed and even activated at temperatures above 1000 K. For practical purposes, Ti 2 N, Zr 2 N, Hf 2 N, V 2 N, Nb 2 N, Ta 2 N are predicted as the best suited materials for CO 2 activation. Moreover, the estimated CO 2 uptake of 2.32–7.96 mol CO 2 ·kg -1 reinforces the potential of these materials for CO 2 abatement.


Introduction
There is no doubt that the continued increase in the concentration of CO2 in the Earth atmosphere directly correlates with global warming and that, in turn, such increase is strongly related to anthropogenic emissions resulting from the use of fossil fuels. 1 The need to reduce the amount of this greenhouse gas has triggered research into CO2 sequestration with the idea not only to contribute decreasing its concentration in the atmosphere 2,3 but of its ulterior utilization as a chemical feedstock in the chemical industry.This may look as a simple idea but its practical implementation into a sufficiently large scale constitutes an enormous scientific and technological challenge. 4This is because of the very high stability of gas phase CO2, the final product of spontaneous combustion of fuels based on organic compounds such as wood, natural gas, or gasoline.
The high chemical stability of CO2 also implies weak interaction with solid substrates, which, according to the well-known Sabatier principle, encounters problems when aiming at using heterogeneous catalysts for CO2 conversion.Hence, in the absence of sufficiently active catalysts, CO2 capture appears as more attractive than direct conversion.Indeed, CO2 capture based on amine-scrubbing technology has become a widely used method. 5,6Nevertheless, because of its non-corrosive nature, low-cost, and easy regeneration, capture by solid materials is considered as advantageous. 7,8The use of solid materials is the basis for the so-called carbon capture and storage (CCS) strategy.The requirements are similar yet less stringent than for heterogeneous catalytic conversion. 9In fact, CCS requires a specific type of solidsubstrate able to adsorb CO2 in a sufficient strong way.Ideally, one would like to have materials able to trigger activation usually involving charge transfer towards the molecule resulting in the generation of bent anionic CO2 δ-species.This is the case for CO2 sequestration through carbonation of natural silicate minerals driving to the formation of carbonates.Unfortunately, this fixes CO2 to the substrate making it unavailable for subsequent conversion. 10,11In order to find appropriate candidates for CCS, a large number of diverse materials has been analysed involving metals, metal oxides, graphene-based materials, zeolites, metalorganic frameworks and transition metal carbides [12][13][14][15][16][17][18][19] although with not completely satisfactory results.
The preceding discussion clearly illustrates the need to go beyond existing materials and to look for alternatives.In this sense, it is worth mentioning that an entirely new family of two-dimensional (2D) transition metal carbides and nitrides has been reported quite recently.These materials have a general Mn+1Xn chemical formula where M stands for early transition metals and X for C or N. 20,21 Depending on n, which runs from 1 to 3, the resulting materials have 3, 5, or 7 atomic layers.Because of the similarities with graphene, including a hexagonal atomic structure, this new materials have been designed as MXenes.
At present, several members of the potentially very large MXene family have been synthesized from the precursor so-called MAX phases. 22This was initially achieved by selectively etching Al from Ti3AlC2 and other Al-containing MAX phases using hydrogen fluoride. 23However, the progress in this field is very fast and fluorine-free synthetic procedures have recently been proposed to synthesize MXenes. 24,25Even, more recently, MXene synthesis from other MAX phases not containing Al has been achieved 26 which shows that the number of these materials already obtained is likely to represent a small fraction of those to be obtained in the near future.
Among other properties recently reviewed, 20,26 MXenes display high surface areas up to 1000 m 2 g - 1 and excellent stability. 23The high surface area displayed by MXenes together to the reactivity of transition metal carbides towards CO2 adsorption and activation 18 triggered a systematic research where the CCS ability of MXenes with M2C formula have been computationally investigated. 27Using adsorption and desorption rates derived from transition state theory and density functional theory based calculations, it has been shown that the M2C MXenes are able to adsorb and release CO2 at rather high temperatures and low partial pressures.Moreover, CO2 uptakes ranging from 2.34 to 8.25 mol CO2 kg -1 of substrate have been predicted which are a direct consequence of their chemical activity and high surface area.It is worth pointing out that this uptake is equivalent to that of other porous material used regularly for CCS purposes such as zeolites, 28,29 derivatives of graphene, 30 or bulk MgO nanopowders 31 .
In the view of the very large number of potential MXene compounds and motivated by the encouraging results obtained for the M2C subset, 27 we analyse the CCS capabilities of the M2N (M= Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, W) family for which information is scarce and new and fascinating chemistry is likely to emerge.In fact, just from the higher electronic conductivity one can already expect that their overall chemistry would be significantly different from that of their carbide counterparts.Moreover, the fact that Ti2N has been already obtained and characterized 32 can be seen as a starting point in the development of efficient synthetic routes leading to other members of the M2N family.

Models
Surface slabs models were used to represent the basal (0001) plan of the M2N materials studied in the present work.The models were built using the crystal structure of the corresponding precursor MAX phase by removing the A element and re-optimizing both lattice parameter and fractional coordinates.The resulting slabs have three atomic layers featuring a M-N-M sandwich-like configuration as schematically shown in Fig. 1.In order to avoid interaction between the adsorbed CO2 molecules in the periodically replicated images a p(3×3) supercell was used containing 18 M and 9 N atoms.Also, since the computational code used (see below) exploits periodic symmetry in the three directions it is necessary to add a vacuum region to avoid the interaction between interleaved slabs; a vacuum width of 10 Å probed to be sufficient to obtain numerically converged results.

Density Functional Theory based calculations
The interaction of CO2 with the M2N(0001) surfaces is studied by means of density functional theory (DFT) 33,34 based calculations within the PBE implementation of the generalized gradient approximation (GGA) for the exchange-correlation potential. 35Additionally, the contribution of dispersion terms to the interaction of CO2 with the M2N(0001) systems is taken into account through the D3 parameterization of the method proposed by Grimme. 36To solve the corresponding Kohn-Sham equations, the valence electron density is expanded in a plane-wave basis set with a cutoff of 415 eV for the kinetic energy, a choice that leads to converged results up to 1 meV whereas the interaction between the valence electron density and the core electrons is accounted for through the projector augmented wave (PAW) method. 37Integration in the reciprocal space is carried out using a 5×5×1 grid of Monkhorst-Pack special k-points. 38Convergence in the geometry optimization is reached when forces acting on nuclei are all below 0.01 eV/Å.The adsorption energy is defined as , where E CO 2 /" # $ corresponds to the energy of CO2 adsorbed on M2N surface and E " # $ stands for the energy of the relaxed pristine M2N surface; E CO 2 corresponds to the energy of an isolated CO2 molecule computed in a symmetric box of 10×10×10 Å dimensions and using the Γ-point only.Finally, ΔEZPE corresponds to difference in zero-point energy (ZPE) of each term contribution calculated in the harmonic approximation.Note that the vibrational frequencies were calculated decoupled form surface phonons and including frustrated rotations/translations.
Attending to this definition of Eads, adsorption is associated to negative values and the more negative the stronger the interaction is.All DFT based calculations are performed with the Vienna ab initio Simulation Package (VASP). 39,40

Adsorption/Desorption Rates
Assuming that CO2 adsorption is a non-activated process, a meaningful choice, the adsorption rate rads can be calculated from the well-known Herz-Knudsen formula as in Eq. ( 1), 18,41 where S0 is the initial sticking coefficient,  :; # corresponds to the CO2 partial pressure above the surface, A stands for the area of an active adsorption site and m corresponds to the mass of the adsorbed molecule.
A conservative value of S0 = 0.40 is selected for our study following a previous analysis where the CO2 capture, storage and activation were investigated on transition metal carbides and two-dimensional M2C MXenes. 18,27Note that equal adsorption probability of all sites (Fig. 1) is assumed and therefore, A is approached as the supercell area of each surface divided by the total number of adsorption sites in it.Three different pressure values representative of different conditions of interest are selected to evaluate rads: i) the atmospheric partial pressure of CO2,  :; # = 40 Pa; 42 ii) a partial pressure of  :; # = 0.15 bar (15•10 3 Pa) which is a reference value for post-combustion exhaust gases; 43 iii) a partial pressure of  :; # = 1.0 bar (10 5 Pa) corresponding to a reference for pure CO2 stream generation from a carbon capture and storage (CCS) system. 44he rate of desorption, rdes, has been estimated from transition state theory (TST) 45 stating that the reaction rate ri of an elementary step is given by where kBT is the product of Boltzmann constant, kB, and the temperature, T, and ΔE would be associated to the zero point corrected energy barrier for the described elementary step, here using the (negative) Eads values implying that the transition state is supposed to be located infinitely close to the desorption final state.In Eq. ( 2), v is the well-known pre-exponential actor term also provided by TST with h, q # and q0 being the Planck constant, and partition functions of the transition and initial states, respectively.Hence one has where vdes contains the partition function of the molecule in an early 2D transition state in the numerator.
This partition function is given by the product ]_ stands for the partition function for translational motion in two dimensions whereas,  \b[ a]_ and  cde a]_ correspond to the rotational and vibrational partition functions of the CO2 molecule in the gas phase, respectively.Once the molecule is adsorbed on the surface all degrees of freedom are considered as vibrations since molecular translations and rotations become frustrated through interaction with the substrate and effectively converted into vibrations.The  cde a]_ partition function contains vibrational contributions only.Finally, note that the electronic partition function is set to 1, which is justified by the high energy of the excited electronic states.
Therefore, the necessary partition functions were evaluated as where vi is the harmonic vibrational frequency of each normal mode as predicted from the present DFT based calculations, either for CO2 molecule in vacuum or adsorbed, 2•Trot is the product of the rotational temperature for CO2 and its symmetry number, 2. Trot is taken from the literature as 0.561 K. 46

Results and discussion
One of the main goals of the present study is to provide qualitative and quantitative information regarding the interaction¾ and possible activation¾ of CO2 with M2N(0001) surfaces.To this end, a systematic computational search has been launched considering different orientations of the CO2 molecule with respect to the surface plane and taking all possible adsorption sites into account.The final atomic structure of the adsorbed CO2 molecules predicted by the geometry optimization using an appropriate DFT based method reduces to eight particular sites and the explored M2N systems display at least one of these configurations.To distinguish among these eight bonding modes a notation is used in which the first digit CBOBOB, and η 3 -CO2-µ 4 -CMOMOB.Note also that whether a given bonding mode appears or not depends on each particular M2N surface.For instance, OBOB and CMCMOB are only found in Ta2N and W2N.
However, on most cases (Hf2N, V2N, Nb2N, Ta2N, Cr2N, Mo2N, and W2N) CB is the most favourable bonding mode for CO2 adsorption.It is worth pointing out that this bonding mode was also found to be the second most favourable one for the interaction of CO2 with the related M2C series of MXenes. 27Note, however, that structural features of the M2N materials here studied lead to sites for CO2 that are more accessible than those of their carbide counterparts.
The analysis of results in Table 1 shows that the PBE-D3 calculated adsorption energy (Eads) values are quite large ranging from -1.03 (Mo2N) to -3.13 eV (Ti2N) indicating a very strong interaction.
Interestingly, these values are only slightly smaller than those reported at the same computational level for their M2C counterparts running from -1.13 to -3.69 eV. 27This origin of this difference is quite simple to understand, the N layer in the M2N systems withdraws more charge density from the metal layers than the C layer in the M2C ones implying a reduction in the charge transfer from the metallic layer of the MXene towards CO2.Table 1 also shows that Eads generally decreases (less negative values indicate weaker adsorption) when moving along the d series, which is consistent with the trends reported for CO2 adsorption on several M3C2 and M2C systems. 18,27,47One remarkable difference between M2C and M2N systems is the importance of the contribution of dispersion terms to Eads.The inclusion of these terms through the D3 Grimme method increases the PBE calculated Eads (in absolute terms) for M2N materials by quite a significant amount (0.7-1.0 eV) where a smaller contribution of ~0.40 eV was found for the studied M2C compounds.One could argue that the larger contribution for M2N materials arises from artefacts in the empirical parameters entering in the D3 methods.To rule out this possibility, calculations have also been carried out with the better theoretically grounded Tkatchenko-Scheffler (TS) 48 or the more sophisticated surface many-body dispersion (MBD) approach 49 .For V2N, the calculated D3 contribution to Eads is -0.69 eV, in full agreement with values of -0.68 and -0.71 eV as predicted from TS and MBD methods, respectively (Additional details regarding these issues can be found in the ESI †).This unexpected important contribution of dispersion terms of up to 1 eV has significant implications in the temperature range at which CO2 desorbs, which we discuss in more detail later.Here we note that, in spite of the strong interaction, the geometry optimization procedure always converges to some of the molecularly adsorbed states described above with no evidence of dissociative chemisorption indicating that this would take place an energy barrier would have to be overcome.Interestingly, a bent adsorbed CO2 molecule is found in most cases with the O-C-O angle in the 111 to 137 º range and the C-O bond distances in the 1.25-1.44Å interval, significantly larger than the value for the isolated gas phase molecule.The chemical activation of adsorbed CO2 is additionally confirmed by the net charged computed using the Bader atoms-in-molecules analysis 50 indicating the formation of highly anionic adsorbed CO2 δ-species with charges running from -0.86 to -2.02 e.The results discussed so far support the claim that M2N systems are appropriate for CO2 activation.
The very large adsorption energy and concomitant activation of CO2 on M2N(0001) surfaces strongly suggest that CO2 may be adsorbed up to quite high temperatures.Making use of the above described theoretical framework rads and rdes rates are estimated for a broad range of temperatures as high as 2000 K.Although the upper limit corresponds to a very high temperature it is below the melting point of the corresponding 3D transition metal nitrides, ranging from 2350 to 3330 K. 51,52 Obviously, the adsorption rate has a non-negligible dependence with the number of CO2 molecules hitting the surface and so on the CO2 partial pressure ( :; # ).Also, desorption rate is determined by the adsorption energy which in turn depends on the bonding mode (Table 1).Hence, two desorption curves for each M2N have been built which differ on whether the PBE or PBE-D3 Eads value is used.This choice allows one to consider the least and most favourable adsorption situations thus avoiding any bias from the choice of a given computational method and providing a reasonable range of values for subsequent analysis.The estimated rads and rdes are illustrated in Fig. 2a for M2N systems with the lowest (Cr2N) and highest (Ti2N) Eads values (see also Tables 1 and S1 in the Electronic supplementary information, ESI †).Note that two sites (CMONON and CNOMOM) with similar adsorption strength are found for Ti2N and, hence, only an estimated rdes including PBE and PBE-D3 is expected for this MXene as reported in Fig. 2a.
Let us now describe in detail the features of the plots of adsorption/desorption rates shown in Fig. 2a.The crossing point between rads and rdes indicates the dynamical equilibrium state where the adsorption and desorption rates coincide, temperatures below the crossing point favour adsorption and, consequently, CO2 becomes sequestered at the material surface.For the case of Cr2N at 40 Pa  :; # (labelled as air), two intersections at temperatures of 106 and 579 K are clearly seen corresponding to the weakest (PBE) and the strongest (PBE-D3) estimates, respectively.In this way, T1-T2 defines the temperature range below which CO2 adsorption dominates and, consequently, it would become stored on the M2N surface, above this temperature desorption will prevail and CO2 will be released to the gas phase.Note that an increase of  :; # to 15•10 3 Pa shifts these limits to the T3-T4 interval corresponding to higher values and this is even further increased at 1 bar pressure as indicated by the T5-T6 interval.Note that these temperature ranges are quite large for Cr2N and W2N and this is mainly due to the significant high contribution of dispersion terms (PBE-D3 value) to the adsorption energy value.The average adsorption/desorption switch temperature for the whole set of M2N surfaces explored in the present work is ~1000 K (see Fig. 2b).Interestingly enough, this appears to be quite below the average value for their M2C counterparts where this occurs at ~1250 K. 27 The above discussion applies to every M2N surface studied and the analysis of the rest of cases is summarized in Fig. 2b where the temperature ranges are shown, independent of the particular adsorption energy and rate.Further information is provided in Table S3 in the ESI † including details regarding the temperature range for each particular site of each M2N surfaces.From Fig. 2b  To further examine the capability of the M2N materials to store CO2 we focus in the total amount of this compound that can be stored CO2 per kg of substrate and compare to the values for different families of materials that have been successfully proved such as Ca-X and 13X zeolites, 28,29 with 3.36 and 2.96 mol CO2 kg -1 , respectively, or derivatives of graphene, e.g.a-RGO-950 30 with 3.36 mol CO2 kg -1 , which are clearly better than using bulk MgO nanopowders, 32 with 0.92 mol CO2 kg -1 .To model the potential amount of CO2 storage by the M2N materials we take into account the surface exposure and possible CO2 loading to the high surface area exposed by 2D MXenes. 23For the p(3×3) supercell model employed in this study, one can safely argue that each M2N surface could simultaneously adsorbed four CO2 molecules.From this quite a conservative estimate one readily predicts that the M2N compounds studied in the present work would be able to adsorb 2.32-7.96mol CO2 kg -1 of substrate (Fig. 3).Clearly, practical operating conditions would require CO2 separation from other competing (combustion) gases prior to adsorption, although recent studies on a series of 2D (MXenes) and 3D transition metal carbides show a high adsorption preference for CO2 with respect to other gases such as CO or CH4 by more than 1.5 eV. 53,54

Conclusions
The computational modelling presented in this study, based on state of the art DFT based calculations and the use of transition state theory derived adsorption/desorption rates highlights the potential of the 2D M2N materials (M = Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, W) for CO2 storage and activation.
Our study reports high adsorption energies up to -3.13 eV accompanied by a noticeably large CO2 activation, evidenced by the appearance of a strongly anionic CO2 (n) is assigned depending on the number of atoms of the CO2 molecule (1, 2, or 3) close enough to the surface (η n -CO2).Clearly, this depends on the orientation of the adsorbed CO2 molecule as clearly seen in the side view in Fig. 1.The notation uses a second digit (m) describing the atomic environment (2, 3, 4, or 5) of the adsorbed CO2 molecule (µ m ) and ends by a specification of the location of the n surface atoms in contact with the CO2 molecule (N, M, or B which correspond to Nitrogen hollow, Metal hollow, or Bridge sites).Hence, the eight different bonding modes are univocally labelled as follow, η 1 -CO2-µ 2 -CB, η 2 -CO2-

Fig. 1 Fig. 2 (
Fig. 1 Side and top view of models featuring CO2 adsorbed on eight possible sites of M2N (0001) surfaces.The colour sequence is as follows: brown and red spheres correspond to carbon and oxygen atoms of the CO2 molecule; and dark and light blue spheres to M upper and bottom layers, respectively, whereas the inner nitrogen layer is represented by dark yellow spheres in the M2N(0001) surface.The coordination notation is given below each site.

Fig. 3
Fig. 3 Predicted CO2 uptake by M2N materials compared to values in the literature for zeolites (Ca-X and 13X), derivatives of graphene (a-RGO-950) and bulk MgO nanopowders is represented by dark blue, red, light blue, and dark yellow bars, respectively.Details are given in the ESI †.

Table 1
δ-adsorbed species with elongated δ(CO) bonds, bent structures, and significant MXene→CO2 charge transfer.The high adsorption energies are remarkable since one must point out that CO2 adsorbs very weakly on most materials.The significantly large adsorption energy value for CO2 on the M2N materials makes them even more suitable than their carbide (M2C) counterparts indicating that M2N materials may constitute effective alternatives for efficient CO2 capture, storage, and activation.This is supported by the temperature ranges for adsorption/desorption switch and by predicted CO2 uptakes ranging from 2.32 to 7.96 mol CO2 kg -1 of substrate, quite competitive to other nowadays-existent material solutions.Finally, the fact CO2 become highly activated suggest that these 2D M2N MXenes could also constitute active catalysts for CO2 conversion although this requires additional research.We hope the present results will stimulate further research in the synthesis of new members of the M2N family as well as experimental studies regarding their CCS capability.Zero point corrected PBE-D3 adsorption energies (Eads) for CO2 above different sites of the M2N 0001) surfaces.PBE values and other further details are given in the ESI †.The coordination notation is denoted in parenthesis to diferenciate the adsorption sites shown in Fig.1.